Service for recognizing and voicing Russian sign language, plus transcription of computer and ambient sounds.
Built at the Sber and UrFU "City Code 300" hackathon and won the "ReadAI Gestures" case
Service features
- Camera video capture
- Gesture-to-text conversion using Sber's SLT2TXT neural network
- Text-to-speech using Sber's Salute Speech
- Computer audio capture and transcription (Salute Speech) with emotion detection: positive, neutral, negative
My role in the project
- Application architecture and microservice interaction design
- Application UI design
- Desktop app development with React + Electron
- Python API for integrating the SLT2TXT neural network